After https://svnweb.freebsd.org/base?view=revision&revision=336593 any make action starting spam errors (svn not work on NFS mounted partitions) svn: E155016: The working copy database at '/usr/src' is corrupt. svn: E155016: The working copy database at '/usr/src' is corrupt. make[2]: "/usr/src/release/Makefile.ec2" line 24: warning: "/usr/local/bin/svn info --show-item last-changed-revision /usr/sr c/release/.." returned non-zero status svn: E155016: The working copy database at '/usr/src' is corrupt. svn: E155016: The working copy database at '/usr/src' is corrupt. make[2]: "/usr/src/release/Makefile.ec2" line 24: warning: "/usr/local/bin/svn info --show-item last-changed-revision /usr/sr c/release/.." returned non-zero status
Do you have the later commits which check if /usr/src/.svn exists before running svn?
Yes last my test install Rev: 337791, error exist. My build server have and export full /usr/src (/usr/obj) On notebook I'm mount /usr/src (exists .svn and /usr/obj) and install Before the resulted commit, the given script worked without errors
Hmm, if we see that we have an SVN checkout, I don't know that there's any good way to determine that we shouldn't be using svn to look at it. Is this a situation where we should simply add "2>/dev/null || true" to hide the errors?
I'm trying to understand exactly what is happening here, because the commit in question should in no way impact the /usr/src checkout. Is the /usr/src export on the NFS host in fact corrupt? Two ways off-hand I can think about that can cause this is: 1) A network interruption during 'svn checkout' or 'svn update' of the tree; 2) concurrent svn commands attempting to manipulate or read from the svn database, for example a 'svn update' from one NFS client while 'svn status' from another NFS client. Can you show what 'svn status' on the NFS exported share from the host looks like? It might be a bit verbose, so the top 20 or so lines might be sufficient.
(In reply to Colin Percival from comment #3) Maybe this will be enough (In reply to Glen Barber from comment #4) On build server simple zfs set sharenfs for /usr/src and /usr/obj On notebook mount_nfs If run any svn/svnlite command on build server all work fine build:/usr/src # svnlite info Path: . Working Copy Root Path: /usr/src .., On notebook x220:/usr/src # svnlite info svn: E155016: The working copy database at '/usr/src' is corrupt. But since no commands are made on the laptop with svn, no problems arose up to this commit. Now, too, there is no particular problem, only that unnecessary error messages are not present
Created attachment 196280 [details] Fix? Does this fix the problem?
(In reply to Colin Percival from comment #6) Yes, it's patch work fine.
Fixed in HEAD (r338141). I'll MFC this at the same time as all the related changes.